But alignment is always going to rely on cooperation of users though? What benefit does the delay offer other than the direct one of a delay?

Why is it going to rely on co-operation if users don't have the means to change the model enough to misalign it?

If we're talking about open-source LLMs, among the best embedding, multimodal, pure and coding LLMs are Chinese (attested and not just benchmarks).